Predicting the Association Between Human Drugs and Targets based on HeteSim Score
Wei Le, Fang Zheng · 2021
In the past decades, drug target prediction has attracted a lot of scholars' attention, and many classic algorithms and models have also been generated. We study from machine learning algorithms and construct a heterogeneous network of drug targets through data processing of biological networks. We construct a drug-drug-like network, a drug-target similar network, and a target-target similar network, and then integrate the above three networks into one heterogeneous network, and select two meta-paths in the heterogeneous network "drug-drug-target" path and the "drug-target-target" path, the normalized HeteSim scores for both paths were calculated, and the HeteSim scores for the two paths were integrated to obtain the final result. The drug-target interaction score (HDTA_HeteSim) is calculated based on different paths in heterogeneous networks, and the algorithm is applied to drug and target prediction. In addition, the AUC (area under the ROC curve) of the HDTA_HeteSim model in the leave-one-out cross-validation has a value of 0.9540, which achieves reliable prediction performance. We also used the DT-Hybrid method and the HDTA_HeteSim method to separately analyze the bromocriptine drug predictions, and found that our algorithm are more efficiency.